1 This manuscript is contextually identical with the following published paper:
1
Specziár A; Árva D; Tóth M; Móra A; Schmera D; Várbíró G; Erős T (2018) 2
Environmental and spatial drivers of beta diversity components of chironomid 3
metacommunities in contrasting freshwater systems. Hydrobiologia, 819, pp 123–143.
4
The original published PDF available in this website:
5
https://link.springer.com/article/10.1007%2Fs10750-018-3632-x 6
7 8
Environmental and spatial drivers of beta diversity components of chironomid 9
metacommunities in contrasting freshwater systems 10
11
András Specziár1,*, Diána Árva2, Mónika Tóth1, Arnold Móra3, Dénes Schmera1, Gábor 12
Várbíró4, Tibor Erős1,5 13
14
1Balaton Limnological Institute, MTA Centre for Ecological Research, Klebelsberg K. u. 3., 15
H-8237 Tihany, Hungary 16
2Research Institute for Fisheries and Aquaculture, National Agricultural Research and 17
Innovation Centre, Anna-liget 8., H-5540, Szarvas, Hungary 18
3Department of Hydrobiology, Institute of Biology, Faculty of Sciences, University of Pécs, 19
Ifjúság u. 6, H-7624, Pécs, Hungary.
20
4Department of Tisza River Research, Danube Research Institute, MTA Centre for Ecological 21
Research, Bem tér 18/C, H-4026, Debrecen, Hungary 22
5Danube Research Institute, MTA Centre for Ecological Research, Karolina u. 29., H-1113, 23
Budapest, Hungary 24
25 26
Corresponding author: Tel.: +36 87448244; email: specziar.andras@okologia.mta.hu 27
2 Abstract Partition of beta diversity into components is a modern method that allows
28
inferences about the underlying processes driving metacommunities. Based on two alternative 29
approaches, we examined the patterns of beta diversity components of chironomids in relation 30
to environmental and spatial gradients in three contrasting freshwater ecosystems. Beta 31
diversity and its replacement component increased from environmentally less heterogeneous 32
lake, through more complex wetland to stream network. Constrained ordination revealed that 33
environmental heterogeneity and spatial processes explain some variation of the patterns of 34
pairwise beta diversity components. Both beta diversity partitioning approaches emphasised 35
the importance of habitat structure and food resource in structuring chironomid 36
metacommunities. However, concurrent approaches provided contrasting results regarding the 37
relative role of underlying mechanisms related to species replacement and richness.
38
Therefore, further research is needed to clarify which of the beta diversity partitioning 39
approaches should be preferred more widely in ecological studies.
40 41
Keywords dispersal, environmental filtering, assemblage, niche-based mechanisms, species 42
richness, species turnover.
43 44
3 Introduction
45 46
Disentangling how and why assemblage composition changes from site to site is fundamental 47
to understand many ecological processes, including principles of metacommunity 48
organization and species coexistence (Leibold et al., 2004; Ricklefs, 2004). This issue is the 49
main research frontier of beta diversity analyses, which received increased interest in the last 50
decades, with many developments in theoretical and analytical grounds (e.g. Dray et al., 2006;
51
Tuomisto, 2010a,b; Anderson et al., 2011; Logue et al., 2011).
52
It has been shown, for example, that pairwise beta diversity measures (i.e. which quantify 53
the differences in the number and identity of species between two sites) can be decomposed 54
into ecologically meaningful components. In fact, two concurring approaches have been 55
elaborated recently to dissect components of differences in assemblages, which are related to 56
the degree of differences in species richness or composition between sites. Baselga (2010, 57
2012; thereafter BAS approach) suggested that beta diversity could be dissected into a species 58
turnover (also termed replacement) and a nestedness resultant component. Sensu BAS the 59
turnover component accounts for the dissimilarity associated with the replacement of some 60
species by others between assemblages and the nestedness resultant component accounts for 61
the dissimilarity associated with species losses in which an assemblage is a strict subset of the 62
other more species rich assemblage. Whereas, Podani & Schmera (2011; POD approach) 63
proposed to decompose beta diversity into species replacement component sensu POD and 64
richness difference component associated with species losses and gains irrespective of 65
nestedness. The species turnover or replacement component in both approaches implies the 66
simultaneous gain and loss of species due to environmental filtering, competition and 67
historical events (Leprieur et al., 2011), and thus reflect the influence of ecological gradients 68
on community structure (Legendre, 2014). Whereas, richness difference including its special 69
4 case, the nestedness, may reflect diversity (number) of ecological niches available at different 70
locations or other processes influencing the number of species (e.g. species introductions and 71
physical barriers) (Legendre, 2014). BAS and POD approaches agree in that for practical 72
purposes the relativized forms of these components should be used. However, it is important 73
to note, that even the relativized species replacement components of the two approaches are 74
calculated differently (although they have the same numerator, but are based on different 75
denominators), and thus, these two measures are neither closely correlated to each other nor 76
could represent the same ecological concept (Legendre, 2014; Baselga & Leprieur, 2015;
77
Podani & Schmera, 2016). Soon after the introduction of pairwise diversity components, their 78
multiple-site versions have also been established both for the BAS (Baselga, 2012) and POD 79
(Ensing & Pither, 2015) approaches.
80
The relative importance of beta diversity components and related measures have been 81
evaluated for several systems and it was concluded that their patterns could be highly variable 82
across taxonomic groups and habitats as well as over time (e.g. Boieiro et al., 2013;
83
Brendonck et al., 2015; Lewis et al., 2016; Alahuhta et al., 2017; Ruhí et al., 2017). Further, 84
recent evaluation of experimental mesocosm data revealed that environmental heterogeneity 85
and dispersal intensity could jointly affect the relative importance of species turnover 86
(replacement) and nestedness resultant components sensu BAS in planktonic 87
metacommunities (Gianuca et al., 2017). However, it is still less known how different 88
environmental and spatial factors influence the relative importance of beta diversity 89
components. Specifically, we do not exactly know whether there are specific environmental 90
and spatial properties which could be more related to a particular component. Revealing the 91
relationship of environmental and/or spatial gradients with these components can help us to 92
better understand the drivers of beta diversity.
93
5 In this study we analyse how the relative importance of components of beta diversity could 94
vary between metacommunities of different ecosystems and in relation to environmental and 95
spatial gradients on the example of chironomids (Diptera: Chironomidae). Chironomids are 96
abundant insects that occur in a wide-range of aquatic habitats and preferred model organisms 97
of freshwater ecological studies. Thanks to their diverse and well-defined species specific 98
environmental requirements chironomids have long been used as indicator organisms in both 99
recent and paleolimnological studies (Brundin, 1958; Sæther, 1979; Gajewski et al., 2005;
100
Milošević et al., 2013; Nicacio & Juen, 2015). Although adults may colonize new habitats 101
rapidly, their flight is generally weak and dispersal happens predominantly passively by winds 102
(Armitage, 1995). Accordingly, chironomid metacommunities are under conjunct control of 103
environmental (i.e. niche-based environmental filtering) and spatial (i.e. dispersal limitation 104
and mass effect) processes even at within lake and wetland scales (Árva et al., 2015a, 2017).
105
However, so far there is only a sole study on the chironomids of spring fens (Rádková et al., 106
2014), which provides some insight into the small scale patterns of their beta diversity 107
components using the POD approach.
108
Specific objectives of the study are: (a) to examine whether the patterns of beta diversity 109
components (i.e. replacement and richness difference sensu POD and turnover (replacement) 110
and nestedness resultant sensu BAS) of chironomid metacommunities contrast in different 111
freshwater systems (i.e. a large and shallow lake, a wetland and a country-wide stream 112
network); (b) to evaluate how these measures are related to between sites differences in 113
various environmental properties (i.e. altitude, catchment, climate, landscape, and local 114
physical-, chemical- and biotic habitat attributes) and spatial distribution of the local 115
assemblages; and (c) to discuss agreement and differences between the results obtained by the 116
two, commonly used, POD and BAS approaches.
117
6 Lake, wetland and stream network ecosystems are major freshwater habitat types, and in 118
general, are under contrasting control of different spatial and environmental processes.
119
Individual lakes generally show moderate environmental heterogeneity most of which 120
concentrated in the littoral zone (Suurkuukka et al., 2012; Árva et al., 2015b) and involve no 121
or little amount of within lake elements acting as dispersal constraints. Wetlands generally are 122
mixtures of aquatic and terrestrial habitats, which exhibit high environmental heterogeneity.
123
Due to their mosaic-like landscape pattern (Gibbs, 2000), dispersal capacity of certain aquatic 124
taxa could be more limited in wetlands compared to lakes. Compared with lakes and 125
wetlands, stream networks may represent the longest environmental gradients, often ranging 126
through elevation and climatic zones. In addition, their dendritic topological structure may 127
inherently restrict dispersal for many organisms (Erős & Campbell-Grant, 2015).
128
Accordingly, for research point (a) we predicted that total beta diversity and its replacement 129
(turnover) component will increase from lake, through wetland to stream network ecosystem 130
due to differences in environmental heterogeneity and dispersal limitation effects between the 131
three freshwater types. For research point (b) we predicted that contribution of relativized 132
species replacement and richness related components to beta diversity will be influenced by 133
both spatial and environmental factors, and the importance of spatial processes will increase 134
along the supposed trend of dispersal limitation from lakes, through wetland to stream 135
network. Finally, since BAS and POD approaches differ in their weighting between processes 136
related to species replacements and richness (Carvalho et al., 2013; Baselga & Leprieur, 137
2015), for point (c) we predicted contrasting results on issues (a) and (b) depending on the 138
approach followed.
139 140
Material and methods 141
Study area 142
7 We used three different freshwater systems for the purpose of this study. These included both 143
lotic and lentic ecosystems, and they differed from each other considerably in their 144
environmental characteristics, habitat complexity and spatial extent. The first is a large and 145
shallow lake (Lake Balaton, Hungary), the second is a wetland (Kis-Balaton, Hungary), while 146
the third is a country-wide stream network system (in Hungary; Fig. 1). Detailed descriptions 147
of these large freshwater systems and maps showing the distribution of sampling sites are 148
available in our recent papers (Árva et al., 2015a, 2017; Erős et al., 2017). Thus we present 149
only a brief comparative description of the systems here.
150
Lake Balaton (46o 42' - 47o 04' N, 17o 15' - 18o 10' E; 104.8 a.s.l.) is a large (593 km2) and 151
shallow (mean depth: 3.2 m) lake. The lake is dominated by homogeneous open water habitat 152
(>85% of the lake area), and consequently most of the environmental heterogeneity and biotic 153
diversity are concentrated in the narrow littoral zone of ca. 200 m width only. Half of the 154
shoreline is covered by reed grass stands, while its remaining part is strongly modified and 155
covered by concrete buildings and ripraps. Small boat harbours situated within the reed grass 156
stand and large sailing vessels and commercial ship harbours bordered by ripraps from waves 157
occur along the whole shoreline and provide special habitats for the biota. In Lake Balaton, 158
128 sites distributed among the characteristic mesohabitats and across the lake area were 159
sampled. Kis-Balaton (46° 34’ - 46° 42’ N, 17° 07’ - 17° 16’ E.; 106 m a.s.l.) is a very 160
shallow (mean depth: <<1 m), lowland wetland area with a total extend of ca. 147 km2. This 161
wetland system is exceedingly heterogeneous with natural and semi-natural aquatic habitats, 162
including large areas with open water, emergent, submerged and floating leaved aquatic 163
macrovegetation, riparian vegetation, wet and inundated forests and meadows, canals either 164
with and without currents, river habitats, ripraps, and separated borrow pits of variable 165
succession stages, as well as extended patches of terrestrial vegetation. In Kis-Balaton, we 166
sampled 79 sites representing the environmental heterogeneity of aquatic habitats and their 167
8 distribution within the system. Whereas, the third study system, the stream network, included 168
51 running water (stream and river) sites, which distributed across the territory of Hungary 169
(range of sites: 46o 6' - 48o 30' N, 16o 12' - 22o 50' E) in the Danube River catchment.
170
Sampling sites were appointed to represent gradients in stream size (mean width: 1.6-186 m;
171
mean depth: 0.015-3.0 m), altitude (from 85 to 261 m a.s.l.) and other influential 172
environmental gradients in climate, landscape, current, substrate characteristics, macrophyte 173
cover and chemical properties in the region.
174 175
Chironomid sampling 176
Benthic chironomid larvae were sampled between 26 June and 13 July 2012 in Lake Balaton 177
and between 23 June and 01 July 2014 in Kis-Balaton. Sediment was sampled by means of 178
Ekman grab and three merged cores taken within a 1 m2 area represented the sample for each 179
site. In addition, surface of stones from riprap habitats in equal area to the Ekman grab 180
samples were cleaned and washed to plastic containers. Both sediment and stone periphyton 181
samples were washed through a 0.25 mm mesh sieve and transported to the laboratory alive in 182
a cooling box. Larvae were separated from sediment by sugar flotation method (Anderson, 183
1959), and then euthanized and stored in 70% ethanol until identification. Stream survey 184
included two sampling occasions in August 2013 and March to April 2014. Chironomid 185
assemblages were assessed according to the multi-habitat sampling protocol proposed by the 186
AQEM project (AQEM Consortium, 2002; Hering et al., 2004). At each site 20 sample units 187
were distributed along a 100 m long stream section to represent proportional area of 188
mesohabitats present. Chironomids were “kick and sweep” sampled using a standard hand net 189
(frame width: 25 cm; mesh size: 1 mm) by the same operator. Samples were preserved and 190
stored in 70% ethanol for laboratory sorting and identification. Chironomid larvae were slide- 191
mounted and identified to species or the lowest possible taxonomic levels.
192
9 193
Habitat assessment 194
Parallel to samplings, we measured series of environmental variables (see Appendix A in 195
Electronic Supplementary Material) that have been found to influence assemblage structure of 196
chironomids in the study region (Árva et al., 2015a,b, 2017; Schmera et al., 2018) and 197
elsewhere (e.g. Real et al., 2000; Rae, 2004; Free et al., 2009; Puntí et al., 2009; Tóth et al., 198
2012). Considered aspects of regional and local environment included groups of variables 199
related to altitude (in streams only), catchment size (in streams only), climate (in streams 200
only), landscape, physical structure of sites, chemical properties of sites, and plants and their 201
remains at sites. Since altitude, catchments size and climate were practically the same for all 202
sites, these variables were not relevant in Lake Balaton and Kis-Balaton studies. Altitude was 203
measured in the field with a GPS device (Garmin Montana 650). Catchment size data were 204
obtained from database of the General Directorate of Water Management of Hungary. Climate 205
variables included mean annual precipitation, number of sunny hours per year and mean 206
annual air temperature data obtained from the CARPATCLIM Database © European 207
Commission - JRC, 2013 (Szalai et al., 2013). Landscape variables for Lake Balaton were the 208
lake basin (i.e. Keszthely-, Szigliget-, Szemes- and Siófok-basin; dummy coded), location 209
along the north-to-south transect of the lake (i.e. northern littoral, offshore and southern 210
littoral; dummy coded), and distances from the closest shore, reed grass stand, floating leaved 211
or submerged macrophyte meadow and open water measured by a GPS device. In Kis- 212
Balaton, landscape variables encompass distances from the closest clump, shore, reed grass 213
stand, floating leaved or submerged macrophyte meadow, and open water. In addition, sites 214
were classified as undisturbed and disturbed, with the latter indicating continuous or recent 215
(i.e. within two years) habitat modifications (e.g. dredging, inundation, vegetation cutting).
216
While, landscape variables for the country-wide stream survey included major land cover 217
10 categories (CLC variables) obtained from the CORINE Land Cover 2006 (European
218
Environmental Agency, 2010) and variables describing bank vegetation (see Appendix A in 219
Electronic Supplementary Material).
220
Local physical, chemical and biotic (plants and organic matter) properties of sites were 221
characterised in a very similar manner in Lake Balaton and Kis-Balaton. At each sampling 222
site, we recorded water depth, Secchi disc depth, current (not relevant in Lake Balaton), 223
temperature and redox potential (not measured in Kis-Balaton) of the uppermost sediment 224
layer, and dissolved oxygen content, pH and conductivity of the water close to the bottom.
225
Emergent, submerged, and floating leaved macrophytes, filamentous algae (Cladophora sp.), 226
moss, riparian vegetation, and tree coverage (%) was estimated visually within a circle of 3 m 227
diameter around the sampling point and the area of the submerged and floating leaved 228
macrophyte stand was recorded by a GPS device and calculated by MapSource version 229
6.16.3. software (Garmin Ltd., www.garmin.com). The substratum of the sites was inspected 230
for percentage compound of clay (grain size ≤0.002 mm), silt (0.002-0.06 mm), sand (0.06-2 231
mm), gravel (2-4 mm), rock (>200 mm), peat, mollusc shells and pure reed grass root 232
(characteristic in some degrading reed grass stands of Lake Balaton). Occurrence of fine 233
(FOM) and coarse (COM) decomposing organic matter particles, reed and tree leaves, and 234
woody debris (excluding leaves) in the sediment, and occurrence of dead trees at the site was 235
rated visually on a six category scale (0-5; where zero denotes absence and 1 to 5 correspond 236
to the 1st to 5th 20% quantiles relative to the maximum observed abundance of that property 237
in the area). Percentage organic matter content was assessed from dry (at 50oC for 72-96 238
hours until constant mass was reached) samples of the upper most 2 cm sediment layer 239
according to the loss-on-ignition method at 550oC for 1 hour (LOI550; Heiri et al., 2001). In 240
addition, chlorophyll-a was extracted from the upper 2 cm sediment layer by hot methanol 241
method (Iwamura et al., 1970) in Lake Balaton, and from whole water column samples by 242
11 acetone method (Aminot & Rey, 2000) in Kis-Balaton, and then, its concentration was
243
measured spectrophotometrically (Shimadzu UV-1601 spectrophotometer).
244
In wadeable streams, 6-15 transects (depending on the complexity of the habitat; Sály et 245
al., 2011) perpendicular to the channel were distributed along each 100 m long sampling 246
section to measure wetted width, and water depth and current velocity (at 60% depth) at 3-6 247
(varied according to the channel width) equally spaced points. In non-wadeable streams and 248
rivers, mean channel width was measured on Google Earth, while current velocity and water 249
depth were averaged from 10-15 measurements along each sampling reach. All the other 250
environmental variables were assessed in the same manner for all type of streams. The 251
substratum of the sites was visually inspected for percentage compound of clay (grain size 252
≤0.006 mm), silt and sand (0.006-2 mm), gravel (2-60 mm), stone (60-400 mm) and rock 253
(>400 mm), as well as for the relative amount of fine (FOM) and coarse (COM) decomposing 254
organic matter particles. Note that these sediment components are not fully equivalent with 255
those applied in lake and wetland systems. Water temperature, conductivity, dissolved oxygen 256
content, and pH were measured with an OAKTON Waterproof PCD 650 portable meter, and 257
concentration of nitrogen (i.e. nitrate and ammonium) and phosphorous (i.e. phosphate and 258
total phosphorous) forms were assessed using Visocolor ECO field kits (Macherey-Nagel 259
GmbH & Co. KG., Germany). Macro- and microalgae (i.e. diatoms; only when they formed 260
visible patches, otherwise they received zero value), emergent, submerged and riparian 261
macrophytes, tree coverage (%) were estimated visually for each sampling section.
262 263
Spatial variables 264
Distribution of sampling sites was modelled by sets of theoretical spatial variables using 265
principal coordinate analysis of among site overland (in air-metres; aPCNM) and watercourse 266
distances (in river-metres; wPCNM; for streams only) according to the modified method of 267
12 Borcard et al. (2004). The relative roles of overland and along watercourse dispersals are not 268
yet fully explored in winged aquatic insects (e.g. Grönroos et al., 2013; Schmera et al., 2018;
269
see also in Discussion), thus we calculated both overland and watercourse distances among 270
the sites of the stream survey. Because these considerations have no or little relevance there, 271
only “overland” geographical distances were used in Lake Balaton and Kis-Balaton. The 272
PCNM variables model the position of each sampling site relative to all the other sites, 273
similarly as they distribute on the map (Borcard et al., 2004; Dray et al., 2006). The procedure 274
we followed to generate PCNM variables however differs in part from the original approach 275
elaborated mainly to identify periodic distance related patterns in the nature (Borcard &
276
Legendre, 2002; Borcard et al., 2004; Dray et al., 2006). Specifically, we did not truncate the 277
distance matrix, but rather used a logarithmic transformation of pairwise distances. The 278
reason of this modification was that we wanted to use spatial variables to model distance and 279
position related dispersal processes with an assumption that the probability of dispersal 280
limitation increases with the geographical distance at a decreasing rate. We believe that 281
logarithmic transformed distance data are more appropriate to capture patterns related to 282
dispersal limitation than distance data truncated according to a subjective distance threshold 283
(e.g. the largest distance between the closest neighbouring sites), and then applying an 284
artificial multiplier for larger distances (e.g. four times the largest distance between the closest 285
neighbouring sites) as originally proposed by Borcard & Legendre (2002). So we constructed 286
matrixes of log(x+1) transformed Euclidean overland and watercourse (in streams only) 287
distances between all pairs of sampling sites obtained from the GPS coordinates and the 288
National GIS Database of Hungary (Institute of Geodesy, Cartography and Remote Sensing, 289
Hungary), respectively, and subjected them to principal coordinate analyses using Past 2.17 290
software (Hammer et al., 2001) to obtain desired sets of PCNM variables. In order to limit the 291
number of potential explanatory variables used in the statistical analysis, we used only the 292
13 first 20 PCNM variables in each data set and excluded all the others with low eigenvalues 293
(<1%), which presumably have little ecological relevance.
294 295
Calculation of beta diversity and its components 296
Here, we briefly summarise the basic algebra of the BAS and POD approaches following 297
Legendre’s (2014) system of symbols. We used the Jaccard index for measuring pairwise 298
similarity (SJ) and 1-SJ for measuring beta diversity (i.e. Jaccard dissimilarity; DJ) among the 299
sampling sites. Beta diversity was further decomposed into relativized additive fractions of 300
species replacement (ReplPJ) and richness difference (RichPJ) components according to the 301
POD method (Eq. 1; Podani & Schmera, 2011), and species replacement (ReplBJ) and 302
nestedness-resultant (NesBJ) components according to the BAS method (Eq. 2; Baselga, 303
2012):
304
𝐷𝐽 = 1 − 𝑆𝐽 =𝑎+𝑏+𝑐𝑏+𝑐 = 𝑅𝑒𝑝𝑙𝑃𝐽+ 𝑅𝑖𝑐ℎ𝑃𝐽 = 2 min(𝑏,𝑐)𝑎+𝑏+𝑐 +𝑎+𝑏+𝑐|𝑏−𝑐| (1) 305
𝐷𝐽 = 1 − 𝑆𝐽 =𝑎+𝑏+𝑐𝑏+𝑐 = 𝑅𝑒𝑝𝑙𝐵𝐽+ 𝑁𝑒𝑠𝐵𝐽 =𝑎+2 min(𝑏,𝑐)2 min(𝑏,𝑐) +𝑎+2 min(𝑏,𝑐)𝑎 ∗𝑎+𝑏+𝑐|𝑏−𝑐| (2) 306
where a is the number of species present in both sites, whereas b and c represent the number 307
of species present only in the first and second, respectively. Equations (1) and (2) can be re- 308
arranged as:
309
1 = 𝑆𝐽+ 𝑅𝑒𝑝𝑙𝑃𝐽 + 𝑅𝑖𝑐ℎ𝑃𝐽 (3)
310
1 = 𝑆𝐽+ 𝑅𝑒𝑝𝑙𝐵𝐽+ 𝑁𝑒𝑠𝐵𝐽 (4)
311
respectively. These relationships summarize the relative amount of similarity (proportion of 312
common species) and difference (beta diversity) related to species replacement and richness 313
difference, and species replacement and nestedness-resultant between the species pools of two 314
sites according to the POD (Eq. 3) and BAS (Eq. 4) approaches, respectively. If these 315
relativized values are calculated for all pairs of sites, then one can analyse components of 316
species level variations in a system including the 2D simplex graphical approach (Podani &
317
14 Schmera, 2011) and relate them to environmental and spatial patterns using constrained
318
ordination and variation partitioning procedures.
319
Although pairwise indexes are good descriptors of between sites patterns across the studied 320
system, but as it has been shown, they cannot account properly for co-occurrence patterns of 321
species in many sites, and thus, may not be ideal tools for comparing whole systems (Diserud 322
& Ødegaard, 2007; Baselga, 2013). Therefore, we also used multiple-site measure of Jaccard 323
dissimilarity and its components to assess the amount of total beta diversity (multiple-DJ) and 324
species replacement (multiple-ReplPJ) and richness difference (multiple-RichPJ) according to 325
the POD approach (Ensing & Pither, 2015), and species replacement (multiple-ReplBJ) and 326
nestedness-resultant (multiple-NesBJ) according to the BAS approach (Baselga, 2012).
327 328
Statistical analysis 329
In order to get more robust data for seasonal stream surveys with many single- and doubleton 330
taxa in the samples, chironomid samples from the two sampling occasions were merged, 331
whereas related environmental data were averaged by sites prior to analyses. Moreover, since 332
pairwise beta diversity partitioning approaches cannot handle zero values, sampling sites 333
without chironomids (zero sites in lake, three in wetland and one in stream network) were 334
excluded from the analyses.
335
We used individual based rarefied (10,000 permutations) taxon richness curves produced 336
with EcoSim 7.72 software (Gotelli & Entsminger, 2011) to compare total (gamma) 337
diversities among the three study systems and to evaluate the adequacy of sampling effort in 338
terms of detection of taxa (Gotelli & Colwell, 2001). To visualise the relationship between the 339
species composition of the three ecosystems and the amount of among sites variation in their 340
metacommunities, we performed non-metric multidimensional scaling (NMDS) analysis for 341
sampling sites based on the Jaccard dissimilarity index with PAST 2.17 software (Hammer et 342
15 al., 2001). In addition, among sites differences in environmental conditions were
343
demonstrated by performing standardized principal component analysis (PCA) for each 344
ecosystem also with PAST 2.17 software (Hammer et al., 2001). These latter results are 345
presented in Appendix B (in Electronic Supplementary Material).
346
Total chironomid beta diversity was assessed by calculating multiple-DJ and its multiple- 347
site components for lake, wetland and stream network ecosystems. Since multiple-site indices 348
might be sensitive to differences in the number of sites sampled (Baselga 2010), we 349
resampled 1000 times the lake and wetland data set to the sample size of stream network 350
(n=50), and calculated the mean and the true 95% confidence interval (CI) of each measure 351
for the resampled data sets. Analyses were performed in R (R Core Team, 2015) using the 352
betapart package (Baselga et al., 2017). The R-script for this analysis is provided in the 353
appendix in Ensing & Pither (2015).
354
Trends of pairwise beta diversity in the three study systems were first visually evaluated 355
using 2D simplex graphical analysis (Podani & Schmera, 2011, 2016) according to the POD 356
and BAS approaches based on equations (3) and (4), respectively. Then, pairwise index 357
values were averaged across all pairs of sites to obtain an alternative percentage 358
decomposition of total beta diversity into its components in each community (Podani &
359
Schmera, 2011; Legendre, 2014). Note that the 2D simplex analysis of beta diversity 360
components have been proposed specifically for the POD approach, and since species 361
replacement and nestedness-resultant components sensu BAS has no meaningful 362
complements, this analysis holds less analytical potential in the BAS approach (Podani &
363
Schmera, 2016). However, to provide some comparative insight into the analytical capacity of 364
the two concurring beta diversity partitioning approaches we show 2D simplex results for the 365
BAS approach as well. Variability of pairwise site scores of each measure was inspected 366
across study systems with permutational one-way analysis of variance (pANOVA; with 999 367
16 permutations) and permutational t post-hoc test performed in R (R Core Team, 2015) using 368
the anova.1way.R and t.perm.R functions written by P. Legendre (available at:
369
http://adn.biol.umontreal.ca/~numericalecology/Rcode/; accessed 05 February 2018).
370
Differences between the two coherent pairwise beta diversity components was analysed in 371
each metacommunities and separately for the POD and BAS approaches with permutational t- 372
test.
373
To evaluate the role of different environmental and spatial (PCNM) variables in the 374
variability of pairwise beta diversity components in the studied chironomid metacommunities, 375
we performed partial direct gradient analysis followed by a variation partitioning approach 376
(Cushman & McGarigal, 2002; Peres-Neto et al., 2006). We run the analyses based on both 377
the POD and BAS approaches and using sites scores from equations (3) and (4) like in the 2D 378
simplex analysis. We preferred this approach over analysing each beta diversity component 379
individually (e.g. via multiple regression or distance based RDA models: Boieiro et al., 2013;
380
Legendre, 2014; Baselga & Leprieur, 2015; Alahuhta et al., 2017) because relativized 381
pairwise beta diversity components and similarity behave similarly, like percentage relative 382
abundances of species. Since these measures sum up to one, their values are not independent 383
from each other and consequently, it could be beneficial to evaluate them collectively. First, 384
we calculated matrixes of between site Euclidean distances for each environmental and spatial 385
variable. These pairwise differences in each specific environmental and spatial variable 386
served then as potential explanatory variables in the multivariate analyses. Of explanatory 387
variables, those measured on continuous scales and representing percentage distribution were 388
log(x+1) and arcsin√x transformed, respectively. Categorical and dummy coded local 389
environmental, pH and spatial PCNM variables were not transformed (see Appendix A in 390
Electronic Supplementary Material). Since preliminary detrended correspondence analysis 391
(DCA) indicated moderate gradient lengths in response variables (i.e. pairwise similarity and 392
17 beta diversity components) for all three study systems and for both POD and BAS approaches 393
(ranging between 1.63-1.90 and 1.84-2.08 in S.D. units, respectively), we chose redundancy 394
analysis (RDA) for further evaluation (Lepš & Šmilauer, 2003). Potential explanatory 395
variables were filtered for collinearity at r>0.7 and subjected to a forward stepwise selection 396
procedure (at P<0.05) in RDA based on Monte Carlo randomization test with 9,999 397
unrestricted permutations. Then, to partition effects of significant variable groups (i.e.
398
altitude, catchment, climate, landscape, physical site properties, chemical site properties, 399
aquatic plants and decomposing organic matter, and spatial) on pairwise beta diversity 400
components of local chironomid assemblages, a series of RDAs and partial RDAs were 401
conducted (Cushman & McGarigal, 2002). DCAs and RDAs were performed using 402
CANOCO version 4.5 software (ter Braak & Šmilauer, 2002).
403 404
Results 405
Gamma and alpha diversities 406
Sampling yielded a total of 13,804 individuals and a system level gamma diversity of 40 taxa 407
(identified at species, species group and genus levels) in lake, 9,321 individuals and gamma 408
diversity of 56 taxa in wetland, and 6,138 individuals and gamma diversity of 120 taxa in the 409
stream network. The cumulative number of observed taxa for the three systems was 157.
410
Proportions of both rare taxa (i.e. single- and doubletons) and taxa with limited distribution 411
(presenting at one or few sites only) were substantial and varied considerably among systems 412
(Appendix C in Electronic Supplementary Material). The number of rare taxa was highest in 413
stream network (16 singletons and 12 doubletons, 13.3% and 10.0% of the total taxa, 414
respectively), intermediate in wetland (nine singletons and two doubletons, 16.1% and 3.6%) 415
and lowest in lake (four singletons and one doubleton, 10.0% and 2.5%). In stream network, 416
29 taxa presented at one site and 23 taxa at two sites only. The same values were 14 and seven 417
18 in wetland, and five and three in lake. Taxon richness per site (alpha diversity) ranged
418
between two and 22 (mean: 8.2; median: 8) taxa in lake, between zero and 25 (6.6; 6) taxa in 419
wetland, and between zero and 35 (14.6; 14) taxa in stream network.
420
Individual based rarefaction analysis also approved highest chironomid gamma diversity in 421
stream network, intermediate gamma diversity in wetland and lowest gamma diversity in lake 422
(Fig. 2). Separation of 95% true CIs of rarefied species richness values among the three 423
systems indicates that these differences are significant (at P<0.05). However, since neither of 424
the rarefaction curves reached a trivial asymptote, it is very likely that more samplings would 425
detect additional taxa in all three systems, especially in stream network.
426 427
Total beta diversity 428
Multiple-DJ indicated extremely high total chironomid beta diversity for the three systems 429
(Table 1). In addition, although the 95% CIs of resampled multiple-DJ separated slightly 430
between lake and wetland systems, differences between the mean multiple-DJ values of the 431
three systems (multiple-DJ=0.969 in lake, 0.976 in wetland and 0.976 in stream network) 432
could be assumed negligible from the practical point of view. More contrasting differences 433
were found between the three systems in the decomposition of beta diversity into its 434
components, especially based on the POD approach (Table 1). Multiple-ReplPJ proved to be 435
highest (0.647) and multiple-RichPJ lowest (0.328) in stream network, while 95% CIs of both 436
measures overlapped between lakes (resampled means: 0.537 and 0.433, respectively) and 437
wetlands (resampled means: 0.513 and 0.433, respectively). On the other hand, the BAS 438
approach counted almost all of the total beta diversity (DJ) to be replacement related 439
component with little or no differences in multiple-ReplBJ values between the three systems.
440
NMDS plot shows that chironomid metacommunity of the stream network had clearly 441
different species composition than metacommunities of the lake and wetland ecosystems (Fig.
442
19 3). This analysis somewhat oppugn the results of multiple-DJ and revealed that the lake
443
chironomid metacommunity could be a nested subset of the wetland fauna with substantially 444
lower internal variability.
445 446
Pairwise beta diversity 447
2D simplex analysis revealed medium to high mean pairwise beta diversity (i.e. low SJ, mean 448
values ranging between 0.163 in wetland and 0.254 in lake; Table 2) in chironomid 449
metacommunities according to the POD approach. Thus pairwise site scores tended to 450
concentrate close to the left side of the ternary diagram, especially in wetland and stream 451
network, but less markedly in lake (Fig. 4a-c). Replacement component of the pairwise beta 452
diversity proved to be slightly more important than the richness difference component in lake 453
and wetland chironomid communities, while in stream network mean replacement was about 454
two times higher than mean richness difference. Mean ReplPJ trended as lake<wetland<stream 455
network, while mean RichPJ was highest in wetland and lowest in stream network.
456
2D simplex analysis under the BAS framework suggested that pairwise beta diversity was 457
clearly dominated by the replacement component in all of the three systems with mean values 458
following a trend of lake<wetland<stream network (Table 2; Fig. 4d-f). Mean NestBJ proved 459
to be similar in lake and wetland, while it was lowest in stream network.
460 461
Environmental and spatial patterns of pairwise beta diversity components 462
In general, the RDA models explained very similar amount of variance in pairwise beta 463
diversity components of chironomids according to the POD and BAS approaches, although 464
the importance of certain explanatory variables and their participation in the final models 465
varied between the two approaches (Table 3; Fig. 5). Total explained variance was lowest in 466
wetland (15.4% and 17.2% in the POD and BAS approaches, respectively), intermediate in 467
20 lake (22.0% and 27.4%) and highest in stream network (25.2% and 24.9% in models with 468
aPCNM, while 31.9% and 29.6% with wPCNM). Pure effect of spatial predictors was 469
negligible (1.5%) in lake, while they explained 3.6-5.4% and 9.9-13.2% of variance in 470
pairwise chironomid beta diversity components in wetland and in stream network, 471
respectively. In stream network, wPCNMs proved to be more effective predictors than 472
aPCNMs based either on their total or pure effect (Fig. 5). On the other hand, pure between 473
site distances were filtered out from all models (i.e. study area × approach type) during the 474
variable selection procedure.
475
Pairwise beta diversity components of chironomid assemblages were more related to 476
environmental than to spatial predictors in all three systems and based on any approaches 477
(Fig. 5). Further, spatial and environmental effects proved to be largely independent as their 478
shared effect remained under 4% in all cases. In lake, environmental variables classified to 479
site physical properties and plants and organic matter groups had the highest predictive power 480
(Table 3). According to the POD approach, increase of richness difference component of beta 481
diversity coincided with increases of between sites differences in distances from the shore, 482
reed and submerged macrophyte stands and in water depth, while species replacement 483
component increased with increasing between sites differences in physical substrate 484
properties, dissolved oxygen concentration, LOI550 and macroalgae coverage (Fig. 6a).
485
Similar tendencies were obtained based on the BAS approach for the nestedness resultant and 486
species replacement components, respectively (Table 3; Fig. 6d). Likewise in wetland, 487
variables belonging to site physical properties and plants and organic matter groups were the 488
most effective predictors of pairwise beta diversity components of chironomids (Table 3).
489
However, the total amount of variance captured by environmental variables was only 490
moderate, especially in the POD approach, and no clear coincidence appeared on the 491
ordination chart between the vectors of beta diversity components and explanatory variables, 492
21 except between replacement component of the BAS and between sites differences in
493
macroalgae coverage, presence of rock, water temperature and conductivity (Fig. 6b,e). In 494
stream network, between sites differences in landscape, site physical, chemical, and plant and 495
organic matter related properties proved to be more or less similarly effective predictors based 496
on their pure effects (Table 3). In this system, increase of richness difference component of 497
the POD approach coincided with increases of between sites differences in concentration of 498
fine decomposing organic matter particles and mean annual air temperature, and decrease of 499
difference in landscape coverage by artificial, non-agricultural vegetation (CLC14) (Fig. 6c).
500
Replacement component of the POD proved to be most related to between sites differences in 501
clay and stone components of the sediment, water current, dissolved oxygen content of the 502
water and catchment area. Very similar environmental patterns were revealed for the 503
nestedness resultant and replacement components of the BAS approach as well (Fig. 6f).
504
Pairwise assemblage similarities correlated negatively with differences in influential 505
environmental properties in all instances (i.e. the less their environments differed the more 506
local assemblages were similar). However, pairwise similarities correlated positively with 507
specific spatial predictors in some cases, specifically in stream network based on the BAS 508
approach and less tightly in wetland based on the POD (Fig. 6).
509 510
Discussion 511
In this study we evaluated metacommunity patterns of chironomids in three different 512
freshwater ecosystems utilizing the quantification tool of beta diversity components. As 513
assumed, the three metacommunities differed largely in their species pools (gamma 514
diversities) and taxa composition. The values of beta diversity, the relative contribution of 515
particular beta diversity components and their relatedness to environmental and spatial 516
variables also differed markedly. The results obtained from different analyses and based on 517
22 concurring beta diversity partitioning approaches (i.e. BAS and POD) also contrasted in some 518
respect.
519
We assumed that environmental heterogeneity increases from individual lake, through 520
individual wetland to country-wide stream network (see Appendix B in Electronic 521
Supplementary Material; site scores are most concentrated in lake and less in stream network 522
ecosystem in PCA plot based on environmental variables), and accordingly, diversity of 523
chironomid metacommunities should increase along the same trend. Although, total species 524
richness (gamma diversity) followed this trend, results about the patterns of beta diversity 525
were less consistent. For example, the multiple-site Jaccard dissimilarity index suggested very 526
similar and extremely high total beta diversity for all three metacommunities, with index 527
values close to their fundamental maximum of one. We consider this result however to be 528
somewhat misleading, which may be related to the weakness of this measure in effectively 529
comparing beta diversity of the samples. Specifically, an index value of one should indicate 530
that all sites are inhabited by completely different composition of species (there are no 531
common species at any two sites). However, this is clearly not the case in our study systems, 532
since the lake metacommunity was represented by only 40 detected taxa for the 128 sites 533
sampled and the wetland metacommunity by 56 taxa for 76 sites, which indicates that many 534
species should be presented at more than one site even at the highest beta diversity possible 535
under such conditions. Nevertheless, means of pairwise Jaccard dissimilarity also indicated 536
high beta diversity for all three metacommunities, but with clear variability among the studied 537
systems. As we hypothesised, mean pairwise beta diversity proved to be the lowest in lake.
538
This system is dominated by open water habitat representing lower environmental 539
heterogeneity compared to the more complex wetland and stream network systems. On the 540
other hand, contrary to our hypothesis wetland metacommunity received higher mean 541
pairwise beta diversity score than stream network metacommunity. We consider that this 542
23 finding may reflect a methodological bias and be related to the higher environmental
543
resolution of point samples in wetland compared to section level samples in streams.
544
We hypothesised that the relative role of the replacement component of beta diversity will 545
increase from lake, through wetland to stream network ecosystem, because higher 546
environmental heterogeneity is likely to favour more intense replacement (turnover) of 547
species from site to site as a result of environmental filtering (species sorting). This 548
assumption was clearly proved based on the pairwise replacement components of the POD 549
and BAS approaches. Whereas, multiple-site replacement component measure (either based 550
on the POD or BAS approach) provided similar scores for lake and wetland. Irrespective of 551
the index type (i.e. multiple-site or pairwise) and the approach (i.e. POB or BAS) used, 552
species replacement was the predominant component of beta diversity in all systems with 553
most marked dominance in stream network. In stream network the high species richness 554
relative to number of sites investigated (120 taxa for 50 sites) resulted more intense species 555
replacement compared to wetland and especially lake ecosystems, which had substantially 556
less species relative to the number of sites. A similar trend in the replacement component 557
relative to species richness was observed in lichen communities by Nascimbene et al. (2013).
558
For aquatic macrophytes, however, Alahuhta et al. (2017) also showed that variation in 559
species composition (i.e. species replacement) primarily accounts for beta diversity in high- 560
diversity regions, while in low-diversity regions richness difference related processes may 561
have noticable role as well.
562
Richness difference component sensu POD and nestedness resultant sensu BAS 563
contributed clearly the least to beta diversity in stream network. Since richness difference is 564
mainly related to variability of number of ecological niches available across sites, it is not 565
surprising that in stream network, where each sample covered wider range of habitats than 566
individual point samples in lake and wetland, received lower scores for these beta diversity 567
24 components. Therefore, variability of number of available niches across sites seemed to be 568
more influential in organizing lake and wetland metacommunities of chironomids with 569
slightly higher pairwise richness difference component scores in wetland. Since chironomids 570
may occur in high diversity along wide ranges of ecological gradients, it is expectable that 571
their metacommunities are more influenced by species replacement, than mechanism related 572
to richness difference (Rádková et al., 2014). However, under extreme environmental 573
conditions their species richness can be very low as well. Our lake and wetland areas included 574
some sites with very low dissolved oxygen concentration and poor food supply, conditions 575
which could be tolerated only by few species, and therefore, these sites could support richness 576
difference related beta diversity. In accordance with our observations, environmental 577
heterogeneity along with the size of the species pool (i.e. gamma diversity) were also 578
identified as the main drivers of pairwise beta diversity components in chironomids at very 579
small spatial scale in spring fens (Rádková et al., 2014). Results on chironomids from 580
different freshwater systems thus also support the fact that regardless of the observed biota, 581
environmental heterogeneity is likely the most important driver of beta diversity 582
(Rosenzweig, 1995; Leibold et al., 2004; Heino et al., 2015).
583
In this study both the POD and BAS approaches supported the conclusion that the 584
contribution of particular beta diversity components to total beta diversity varied substantially 585
among the three systems. However, results obtained based on the two approaches are not in 586
full agreement in that how chironomid beta diversity is organized. Namely, as it had been 587
shown earlier, the BAS approach gives more weight to the species replacement component 588
than the POD approach (Carvalho et al., 2013; Baselga & Leprieur, 2015) and this difference 589
is apparent in this study as well (Table 1 and 2). Nevertheless, the predominant contribution 590
of the replacement component in all three systems was consistently indicated by both 591
25 approaches, which suggests that niche based processes (species sorting) could play a major 592
role in organising chironomid metacommunities (Cottenie; 2005; Van der Gucht et al., 2007).
593
Concerning the outstanding role of environmental heterogeneity in metacommunity 594
processes (Leibold et al., 2004; Heino et al., 2015), it is not surprising that its effect could also 595
be captured in relative patterns of pairwise beta diversity components in all three chironomid 596
metacommunities using both the POD and BAS approaches. This finding supports that 597
environmental heterogeneity influences not only the variability of local assemblages, but it 598
also affects the relative roles of underlying mechanisms related to species replacement and 599
richness difference. Replacement and richness difference or nestedness resultant components 600
of beta diversity are influenced by different ecological processes and thus generally relate to 601
different environmental and spatial attributes as well (e.g. Boieiro et al., 2013; Legendre, 602
2014; Lewis et al., 2016; Gianuca et al., 2017). Below, we give several examples how 603
components of pairwise beta diversity can be associated with different environmental and/or 604
spatial gradients in the studied systems.
605
In the studied lake system, most chironomid taxa are associated with the littoral zone, 606
while the offshore area is quite species poor (Árva et al., 2015a). Therefore, it is not 607
surprising that vectors of the richness difference component of the POD and nestedness 608
resultant component of the BAS approaches coincided with between site differences in water 609
depth and variables representing distances from particular elements of the littoral zone (e.g.
610
distances from the shore line, reed grass stand and submerged macrovegetation) in the RDA 611
plot (Fig. 6a,d). On the other hand, the role of replacement component either using the POD 612
or BAS approach increased with between site differences of environmental attributes that 613
proved to be important to differentiate between the four main chironomid assemblage clusters 614
in the lake, such as: (1) northern macrophyted littoral and sheltered boat harbours with silt 615
sediment and high LOI550, (2) ripraps (rocks) with algal coating, (3) open water with silt 616
26 sediment and low LOI550, and (4) southern littoral with sand sediment and low LOI550 617
(Árva et al., 2015a). The high congruency in response of species distribution patterns and beta 618
diversity components to environmental gradients could be owing to markedly separated 619
habitat types and related ecological processes in Lake Balaton. In the studied wetland, both 620
micro- and meso-scale environmental heterogeneity is so high that neither habitats nor 621
chironomid assemblages form clear clusters (Árva et al., 2017). This diverse patterning and 622
probable complexity of the underlying ecological mechanisms could be the reason why 623
relative importance of beta diversity components did not provide clear relationship with the 624
considered environmental predictors. Moreover, the only clear congruence between the POD 625
and BAS approaches was that increasing replacement was associated with the difference in 626
presence of rock at compared sites (Fig. 6b,e). Rocks placed to some flow exposed sections 627
represent unique, artificial habitats in this system. Since rocks have dense algal coating and 628
consequently better oxygen supply than other substrates, they are inhabited by chironomid 629
taxa which are not characteristic in other habitats of this wetland area (Árva et al., 2017).
630
Further, in wetland, richness difference component of the POD approach tended to increase 631
with increasing difference in water depth between the sites (Fig. 6b) due to the lower number 632
of chironomid taxa in deeper habitats. This is likely in response to lower number of ecological 633
niches in the deeper and less heterogeneous open water environment similarly to lake. In 634
stream network, richness difference component of the POD and nestedness resultant 635
component of the BAS approach were associated with increasing difference in the ratio of 636
fine particle decomposing organic matter in the sediments (Fig. 6c,f). In addition replacement 637
component was associated with differences in a series of environmental properties like 638
sediment physical structure, stream width and dissolved oxygen content in both approaches.
639
Overall these findings indicate that a multitude of environmental gradients influence patterns 640
of species replacements and richness difference or nestedness resultant components of beta 641
27 diversity in chironomid metacommunities. This patterning is in agreement with relative
642
abundance based constrained assemblage patterns in the region (Árva et al., 2015a, 2017;
643
Schmera et al., 2018) and emphasises the prominent role of habitat structure and range of food 644
resource in the organization of chironomid metacommunities.
645
Components of beta diversity may be structured spatially even besides the effect of 646
spatially structured environmental filters. For instance, Boieiro et al. (2013) identified strong 647
pure spatial effect in both the replacement and richness difference components of POD when 648
examined the beta diversity of ground beetles in Madeira Island Laurisilva. Carvalho &
649
Cardoso (2014) provided another example of how the components of beta diversity change 650
with dispersal possibilities. They revealed that variation in community composition of spiders 651
was related mainly to replacement in case of good dispersers and to richness difference in 652
dispersal-limited taxa using POD. In the latter group, geographical distance was an important 653
predictor of between community dissimilarity (beta diversity). In our study systems spatial 654
effect was the least important in lake, where the dominance of open water habitat enables 655
relatively free dispersal for flying imagos. Further, the unique environmental conditions in the 656
littoral zone favour an efficient environmental filtering and also antagonize potential 657
colonization of abundant open water species. On the other hand the heterogeneous landscape 658
of wetland including also variable areas of terrestrial habitats and unevenly distributed 659
patches of tall trees and clamps may represent spatially structured dispersal constraints for 660
chironomids (Delettre et al., 1992), and result a more pronounced spatial structure in pairwise 661
assemblage composition relationships as well (c.f. Kärnä et al., 2015). Whereas, the country- 662
wide stream network system covers the largest area and the most heterogeneous landscape, 663
therefore it is not surprising that this metacommunity proved to be most structured spatially.
664
There is a yet not fully disentangled variability in dispersal of different macroinvertebrate 665
groups in concern to what extent their movement happens overland or along watercourse 666
28 (Grönroos et al., 2013; Kärnä et al., 2015; Schmera et al., 2018). Here we obtained a better 667
explanatory power for along water course spatial predictors (wPCNMs) than for predictors 668
defined based on overland distances (aPCNMs) for beta diversity patterns in stream network.
669
Although there are indications that dispersal of chironomids and several other flyable aquatic 670
macroinvertebrates may be more confined to movement along the watercourse in habitats 671
bordered by tall forest vegetation, in general these organisms are known to disperse quite 672
effectively overland as well (Delettre et al., 1992; Armitage, 1995; Delettre & Morvan, 2000).
673
On the other hand, in streams eggs and larvae of chironomids are also distributed by the water 674
current (Pinder, 1995), which may emphasize the importance of watercourse distribution over 675
overland dispersal. In fact, further research is needed to evaluate whether this observed 676
pattern has a valid background from dispersal behaviour of chironomids or not. Since 677
environmental properties themselves are often spatially structured, it is not rare that identified 678
environmental and spatial effects overlap as well (Gilbert & Bennett, 2010; Legendre &
679
Legendre, 2012). However, results of the variation partitioning prove that in our systems the 680
overlap between the identified environmental and spatial effects is only moderate.
681 682
Conclusions 683
We demonstrated that both beta diversity and its replacement component increased in 684
chironomid metacommunities from environmentally less heterogeneous lake, through more 685
complex wetland to extended stream network ecosystem. Results proved that the relative role 686
of metacommunity assembly mechanisms related to species replacement and richness 687
difference or nestedness resultant components of beta diversity could also vary substantially 688
across ecosystems in chironomids. We found that environmental heterogeneity and spatial 689
processes explain some variation of the patterns of pairwise beta diversity components in 690
chironomid metacommunities, and the most influential environmental attributes in this regard 691
29 could be the habitat structure and the range of food resource. However, the wider applicability 692
of beta diversity components is still hampered by the limits of particular indexes and the 693
discrepancies between the results of concurrent approaches. Given the substantial differences 694
between the interpretations of species replacement by the POD and BAS approaches, further 695
research is needed to clarify which of the approaches should be preferred to assure general 696
comparability over a wide-range of studies.
697 698
Acknowledgements 699
We thank Endre Bajka, Pál Boda, Gabriella Bodnár, Máté Bolbás, Tamás Bozoki, András 700
Csercsa, Eszter Krasznai, Attila Mozsár, Adrienn Tóth for their assisstance in the field. This 701
research was supported by the OTKA K104279. The work of Mónika Tóth was also 702
supported by the János Bolyai Research Scolarship of the Hungarian Academy of Sciences.
703 704
References 705
Alahuhta, J., S. Kosten, M. Akasaka, D. Auderset, M. Azzella, R. Bolpagni, C. P. Bove, P. A.
706
Chambers, E. Chappuis, J. Clayton, M. de Winton, F. Ecke, E. Gacia, G. Gecheva, P.
707
Grillas, J. Hauxwell, S. Hellsten, J. Hjort, M. V. Hoyer, C. Ilg, A. Kolada, M. Kuoppala, 708
T. Lauridsen, E. H. Li, B. A. Lukács, M. Mjelde, A. Mikulyuk, R. P. Mormul, J.
709
Nishihiro, B. Oertli, L. Rhazi, M. Rhazi, L. Sass, C. Schranz, M. Søndergaard, T.
710
Yamanouchi, Q. Yu, H. Wang, N. Willby, X. K. Zhang & J. Heino, 2017. Global 711
variation in the beta diversity of lake macrophytes is driven by environmental 712
heterogeneity rather than latitude. Journal of Biogeography 44: 1758-1769.
713
Aminot, A. & F. Rey, 2000. Standard procedure for the determination of chlorophyll-a by 714
spectroscopic methods. International Council for the Exploration of Sea, Denmark.
715
30 Anderson, R. O., 1959. A modified flotation technique for sorting bottom fauna samples.
716
Limnology and Oceanography 4: 223–225.
717
Anderson, M. J., T. O. Crist, J. M. Chase, M. Vellend, B. D. Inouye, A. L. Freestone, N. J.
718
Sanders, H. V. Cornell, L. S. Cornita, K. F. Davies, S. P. Harrison, N. J. B. Kraft, J. C.
719
Stegen & N. G. Swenson, 2011. Navigating the multiple meanings of diversity: a 720
roadmap for the practicing ecologist. Ecology Letters 14: 19–28.
721
AQEM Consortium, 2002. Manual for the application of the AQEM system. A 722
comprehensive method to assess European streams using macroinvertebrates, developed 723
for the purpose of the Water Framework Directive. Version 1.0.
724
Armitage, P. D., 1995. Behaviour and ecology of adults. In Armitage, P. D., P. S. Cranston &
725
L. C. V. Pinder (eds), The Chironomidae: Biology and ecology of non-biting midges.
726
Chapman & Hall, London: 194–224.
727
Árva, D., M. Tóth, H. Horváth, S. A. Nagy & A. Specziár, 2015a. The relative importance of 728
spatial and environmental processes in distribution of benthic chironomid larvae within 729
a large and shallow lake. Hydrobiologia 742: 249–266.
730
Árva, D., A. Specziár, T. Erős & M. Tóth, 2015b. Effects of habitat types and within lake 731
environmental gradients on the diversity of chironomid assemblages. Limnologica 53:
732
26–34.
733
Árva, D., M. Tóth, A. Mozsár & A. Specziár, 2017. The roles of environment, site position, 734
and seasonality in taxonomic and functional organization of chironomid assemblages in 735
a heterogeneous wetland, Kis-Balaton, Hungary. Hydrobiologia 787: 353–373.
736
Baselga, A., 2010. Partitioning the turnover and nestedness components of beta diversity.
737
Global Ecology and Biogeography 19: 134–143.
738
Baselga, A., 2012. The relationship between species replacement, dissimilarity derived from 739
nestedness, and nestedness. Global Ecology and Biogeography 21: 1223–1232.
740